Table 2 Results of binary cross entropy loss function with different weights
From: Scale-aware dense residual retinal vessel segmentation network with multi-output weighted loss
foreground weight | background weight | dice | acc | mIoU | recall |
|---|---|---|---|---|---|
0.9 | 0.1 | 0.8027 | 0.9665 | 0.8175 | 0.8815 |
0.8 | 0.2 | 0.8032 | 0.9665 | 0.8180 | 0.8805 |
0.7 | 0.3 | 0.8004 | 0.9662 | 0.8156 | 0.8797 |
0.6 | 0.4 | 0.7973 | 0.9659 | 0.8134 | 0.8786 |
0.5 | 0.5 | 0.7950 | 0.9656 | 0.8116 | 0.8766 |